CLASSIFICATION AND RE-IDENTIFICATION
Aspects of the present disclosure include methods, systems, and non-transitory computer readable media that perform the steps of receiving one or more snapshots, extracting one or more features from the one or more snapshots, and providing the one or more features to a first classification layer for classifying a first target and a second classification layer for re-identifying a second target.
1 . A neural network, comprising:
feature layers configured to:
receiving one or more snapshots;
extracting one or more features from the one or more snapshots; and
providing the one or more features to a first classification layer and a second classification layer;
the first classification layer configured to re-identify a first target; and
the second classification layer configured to classify a second target.
2 . The neural network of claim 1 , wherein:
the first classification layer is further configured to transmit a target identification label associated with the first target to a first classification error component; and
the second classification layer is further configured to transmit a class label associated with the second target to a second classification error component.
3 . The neural network of claim 2 , further comprising the first classification error component, wherein the first classification error component is configured to:
receive a ground truth target identification; and
generate a target identification error based on the target identification label and the ground truth target identification.
4 . The neural network of claim 3 , further comprising the second classification error component, wherein the second classification error component is configured to:
receive a ground truth class; and
generate a class error based on the class label and the ground truth class.
5 . The neural network of claim 4 , further comprising a combined error component configured to:
receive the target identification error and the class error;
generate a combined error based on the target identification error and the class error;
generate one or more updated parameters based on the combined error; and
transmit the one or more updated parameters to at least one of the feature layers, the first classification layer, or the second classification layer.
6 . The neural network of claim 5 , wherein generating the one or more updated parameters comprises generating the one or more updated parameters for minimizing the combined error.
7 . The neural network of claim 1 , further comprising a third classification layer configured to re-identify a third target.
8 . The neural network of claim 7 , wherein:
the first classification layer is further configured to transmit a first target identification label associated with the first target to a first classification error component;
the second classification layer is further configured to transmit a class label associated with the second target to a second classification error component; and
the third classification layer is further configured to transmit a second target identification label associated with the third target to a third classification error component.
9 . The neural network of claim 8 , further comprising:
the first classification error component configured to:
receive a first ground truth target identification; and
generate a first target identification error based on the first target identification label and the first ground truth target identification;
the second classification error component configured to:
receive a ground truth class; and
generate a class error based on the class label and the ground truth class; and
the third classification error component configured to:
receive a second ground truth target identification; and
generate a second target identification error based on the second target identification label and the second ground truth target identification.
10 . The neural network of claim 9 , further comprising a combined error component configured to:
receive the first target identification error, the class error, and the second target identification error;
generate a combined error based on the first target identification error, the class error, and the second target identification error;
generate one or more updated parameters based on the combined error; and
transmit the one or more updated parameters to at least one of the feature layers, the first classification layer, or the second classification layer.
11 . The neural network of claim 1 , wherein the feature layers are further configured to provide the one or more features to the first classification layer and the second classification layer simultaneously, contemporaneously, or in parallel.
12 . The neural network of claim 1 , wherein the feature layers comprises at least one of a transform operator, a pattern detector, a convolution operator, or a filter.
13 . The neural network of claim 1 , wherein the first target and the second target are the same or different.
14 . A non-transitory computer readable medium comprising instructions stored therein that, when executed by a processor of a system, cause the processor to:
cause feature layers to:
receive receiving one or more snapshots;
extracting one or more features from the one or more snapshots; and
providing the one or more features to a first classification layer and a second classification layer;
cause the first classification layer configured to re-identify a first target; and
cause the second classification layer configured to classify a second target.
15 . The non-transitory computer readable medium of claim 14 , further comprising instructions stored therein that, when executed by the processor of the system, cause the processor to:
cause the first classification layer to transmit a target identification label associated with the first target to a first classification error component; and
cause the second classification layer to transmit a class label associated with the second target to a second classification error component.
16 . The non-transitory computer readable medium of claim 15 , further comprising instructions stored therein that, when executed by the processor of the system, cause the processor to cause the first classification error component to:
receive a ground truth target identification; and
generate a target identification error based on the target identification label and the ground truth target identification.
17 . The non-transitory computer readable medium of claim 16 , further comprising instructions stored therein that, when executed by the processor of the system, cause the processor to cause the second classification error component to:
receive a ground truth class; and
generate a class error based on the class label and the ground truth class.
18 . The non-transitory computer readable medium of claim 17 , further comprising instructions stored therein that, when executed by the processor of the system, cause the processor to cause a combined error component to:
receive the target identification error and the class error;
generate a combined error based on the target identification error and the class error;
generate one or more updated parameters based on the combined error; and
transmit the one or more updated parameters to at least one of the feature layers, the first classification layer, or the second classification layer.
19 . A method of classification and re-identification, comprising:
receiving one or more snapshots;
extracting one or more features from the one or more snapshots; and
providing the one or more features to a first classification layer for classifying a first target and a second classification layer for re-identifying a second target.
20 . The method of claim 19 , wherein extracting the one or more features comprises extracting the one or more features using a plurality of feature layers.